How do I install the Source Verify skill?

Run npx skillmds@latest add frankxai/source-verify in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Source Verify safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Source Verify?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Source Verify free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Source Verify?

frankxai (@frankxai) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Source Verify by frankxai — Agent Skill

Name: Source Verify
Author: frankxai

Source Verify

Use to verify every factual claim in a Blue Life Commons draft is backed by an acceptable source before PR. Defends against AI-generated text drifting from cited fact. Triggers — "verify sources", "check citations", "is every claim sourced".

frankxai 600d7c8 1.7 KB Updated Sep 17, 2026

File contents

Source verification (claim-level)

The #1 integrity failure for an agent-authored commons is plausible-but-uncited sentences slipped between the real citations. This skill catches them.

Procedure

Extract every factual sentence (a claim about the world: distribution, diet, behavior, population, status, law, statistics). Skip clearly-labeled first-person field observations and pure logistics.
For each, identify the supporting source in the artifact's sources[] / Sources section.
Check the source tier against SOURCES.md:
- Tier 1 (peer-reviewed, IUCN assessment, government monitoring) — required for core science claims.
- Tier 2 (NGO/museum/university outreach) — acceptable with attribution.
- Tier 3 (science journalism, documentaries) — context only, never a core claim.
- Never: forums, AI output, uncited blogs, social media.
Where the commons uses marine-mcp, call lookup_source to confirm a cited URL is real and already known to the corpus.

Output

SOURCES: PASS, or a list of un-sourced / under-tiered claims:

- "<sentence>" — no source / Tier-3 used for a core claim → add Tier 1–2 or remove the claim.

If any claim cannot be sourced, the rule is: remove the claim or mark the artifact status: needs-sources. Never leave an unsourced factual sentence in a merged artifact.

Built on SIP · Blue Life Commons.

Frequently asked questions